Noncausal two-stage image filtration at presence of observations with anomalous errors

Abstract

Introduction. It is necessary to develop adaptive algorithms, which allow to detect such regions and to apply filter with respective parameters for suppression of anomalous noises for the purposes of image filtration, which consist of regions with anomalous errors. Development of adaptive algorithm for non-causal two-stage images filtration at pres-ence of observations with anomalous errors. The adaptive algorithm for noncausal two-stage filtration is developed. On the first stage the adaptive one-dimensional algorithm for causal filtration is used for independent processing along rows and columns of image. On the second stage the obtained data are united and a posteriori estimations are calculated. Results of experimental investigations. The developed adaptive algorithm for noncausal images filtration at presence of observations with anomalous errors is investigated on the model sample by means of statistical modeling on PC. The image is modeled as a realization of Gaussian-Markov random field. The modeled image is corrupted with uncorrelated Gaussian noise. Regions of image with anomalous errors are corrupted with uncorrelated Gaussian noise which has higher power than normal noise on the rest part of the image. Conclusions. The analysis of adaptive algorithm for noncausal two-stage filtration is done. The characteristics of accuracy of computed estimations are shown. The comparisons of first stage and second stage of the developed adaptive algorithm are done. Adaptive algorithm is compared with known uniform two-stage algorithm of image filtration. According to the obtained results the uniform algorithm does not suppress anomalous noise meanwhile the adaptive algorithm shows good results.